Ethan Marsh
@ryebreadreturns.bsky.social
Yet another twitter refugee
Now, given our inability to stop AI from lying right now, when AI is comparatively simple, how much faith do you have in our ability to stop much more complex and advanced models from lying to us?
October 16, 2025 at 4:20 PM
Now, given our inability to stop AI from lying right now, when AI is comparatively simple, how much faith do you have in our ability to stop much more complex and advanced models from lying to us?
Not because there's no physical way to "pull the plug". There probably will be. But because, in practice, we won't be able to justify making the decision to pull the plug. At that point, our only hope for survival is that the AI never decides to lie to us about what's in our best interests.
October 16, 2025 at 4:20 PM
Not because there's no physical way to "pull the plug". There probably will be. But because, in practice, we won't be able to justify making the decision to pull the plug. At that point, our only hope for survival is that the AI never decides to lie to us about what's in our best interests.
Of course not. That's the future we're headed for, except replace "chess" with "everything". One day, AI will get so good that CEOs won't be able to justify to shareholders any decision that goes against the AI. And that is the point at which humans have effectively "lost control".
October 16, 2025 at 4:20 PM
Of course not. That's the future we're headed for, except replace "chess" with "everything". One day, AI will get so good that CEOs won't be able to justify to shareholders any decision that goes against the AI. And that is the point at which humans have effectively "lost control".
Why? Because when AI is good enough, it will be impossible for humans to "beat" its advice, in the same way it has become impossible to for humans to "beat" AI at chess. If millions of lives or billions of dollars hung in the balance, would you ever put your money in a human chess player vs AI?
October 16, 2025 at 4:20 PM
Why? Because when AI is good enough, it will be impossible for humans to "beat" its advice, in the same way it has become impossible to for humans to "beat" AI at chess. If millions of lives or billions of dollars hung in the balance, would you ever put your money in a human chess player vs AI?
The risk is not robots physically dominating us. The risk is humans turning their decision-making process over to AI.
It will start with humans just asking AI what it thinks about a situation. But when it's good enough, we'll find that we won't be able to go against the AI's recommendations.
It will start with humans just asking AI what it thinks about a situation. But when it's good enough, we'll find that we won't be able to go against the AI's recommendations.
October 16, 2025 at 4:20 PM
The risk is not robots physically dominating us. The risk is humans turning their decision-making process over to AI.
It will start with humans just asking AI what it thinks about a situation. But when it's good enough, we'll find that we won't be able to go against the AI's recommendations.
It will start with humans just asking AI what it thinks about a situation. But when it's good enough, we'll find that we won't be able to go against the AI's recommendations.